US-Based AI Data Insight Tool -- 2

The goal is to turn a stream of numerical records stored in my production database into clear, actionable insights through an end-to-end AI solution. Because the data is already structured and continuously updated, the work spans the full stack: secure data ingestion, feature engineering, model development and evaluation, plus a lightweight interface that lets non-technical stakeholders query results in real time.

You’ll work exclusively with numerical data pulled directly from the database layer, so strong SQL fluency is essential alongside your machine-learning toolkit (Python, scikit-learn, TensorFlow / PyTorch, pandas, NumPy). I also need you comfortable standing up a small front-end—React or similar is fine—and exposing the models behind a RESTful (or GraphQL) API. Containerising everything in Docker and wiring in CI/CD on GitHub Actions will round out the build.

Deliverables
1. ETL pipeline that syncs live database tables into an ML-ready data store
2. Clean, documented feature engineering code with reproducible notebooks or scripts
3. Trained models that surface trends, anomalies, and KPI forecasts with measured performance metrics (accuracy, MAE, or RMSE as applicable)
4. API endpoints returning predictions and explanatory statistics
5. Minimal web dashboard for interactive exploration of results
6. Docker compose file and deployment notes so the whole stack spins up locally or in the cloud without guesswork
7. Concise developer README plus a one-pager for business users explaining how to read the insights

This project is open only to full-stack AI/ML developers physically located in the United States; please confirm your state of residence when you reply. If you’re ready to own the pipeline from data pull to polished output, I’m eager to get started.